Research On Query Log Data Of Civil Aviation For Passenger Demand Modeling

In modern civil aviation industry, passenger demand forecast is the core issue of revenue management for airline. Accurate demand model can help airlines better develop marketing strategies, reduce costs and increase profitability. The traditional demand modeling method rely on historical passenger data, its immediacy and sensitivity have been limited by the data source. Therefore, finding a new modeling idea is a very valuable topics for civil aviation industry.In this paper, we commence with the innovation that Internet has brought to the civil aviation industry, discovering that ticket information query over Internet become increasingly popular before passengers’ travel. A large number of passengers query log data is saved in the civil aviation GDS (Global Distribution System) which contains a wealth of valuable information. We develop a deep analysis and mining to the travelers query log data, provide a new method for civil aviation demand modeling.Step one, we conducted a survey and analysis to the query channel system of GDS, found that different types of querying body exhibit different behavior patterns, also their demand behind the query is different. Therefore, we define a query channel behavior patterns category, and then were constructed channel query behavior characteristic set of attributes from the query time-frequency and content distribution. Finally, we build a channel behavior multi-classification methods.Step two, we collected a large amount of historical data include tickets, boarding and revenue to exposit the real demand, and experimental verify the relevance between query behavior and real demand of passengers. Finally, we classify the query data as input, replacing the historical passenger data in traditional prediction algorithms, build up a new demands index model.We experiment with both traditional prediction algorithms and our new method in different scenario include:normal day, holidays and regionalism event. The results by comparison show that new method based on query data performance better during holidays and regionalism events, and the demand index is closer to the real passenger flow volume.